File size: 11,087 Bytes
2955ecc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
schema_version: '0.1'
created: '2026-05-16'
purpose: Public source ledger for Sxx source_refs used by observable rule files.
source_policy:
  source_refs_scope: Rules under observables/rules use these source keys for public provenance.
  self_contained_public_ledger: true
sources:
- id: S01
  title: EU AI Act Article 51
  url: https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-51
  summary: systemic-risk GPAI presumption at greater than `10^25` FLOP, with threshold update provisions.
- id: S02
  title: EU AI Act Article 52
  url: https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-52
  summary: notification timing after a systemic-risk GPAI model meets the Article 51 condition.
- id: S03
  title: EU AI Act Annex XI
  url: https://ai-act-service-desk.ec.europa.eu/en/ai-act/annex-11
  summary: technical documentation includes computational resources, training time, and energy consumption.
- id: S04
  title: NVIDIA H100 Tensor Core GPU official product/spec pages
  url: https://www.nvidia.com/en-us/data-center/h100/
  summary: advertised peak tensor rates, memory capacity, memory bandwidth, NVLink, and power envelopes.
- id: S05
  title: NVIDIA H200 Tensor Core GPU official product/spec pages
  url: https://www.nvidia.com/en-us/data-center/h200/
  summary: memory capacity, HBM bandwidth, NVLink, and tensor performance context.
- id: S06
  title: NVIDIA Blackwell/B200 official product/spec pages
  url: https://www.nvidia.com/en-us/data-center/technologies/blackwell-architecture/
  summary: advertised generation-level performance and memory/fabric context.
- id: S07
  title: NVIDIA DCGM Exporter docs
  url: https://docs.nvidia.com/datacenter/dcgm/latest/gpu-telemetry/dcgm-exporter.html
  summary: Prometheus export of GPU telemetry such as power, energy, temperature, clocks, memory,
    PCIe/NVLink, profiling counters, and MIG-aware metrics.
- id: S08
  title: NVIDIA DCGM field IDs
  url: https://docs.nvidia.com/datacenter/dcgm/latest/dcgm-api/dcgm-api-field-ids.html
  summary: field definitions for utilization, power, temperature, clocks, errors, NVLink/PCIe, profiling,
    and health counters.
- id: S09
  title: NVIDIA System Management Interface docs
  url: https://docs.nvidia.com/deploy/nvidia-smi/
  summary: queryable device inventory, memory, utilization, clocks, power, temperature, performance
    states, throttle reasons, ECC, retired pages, accounting, and MIG state.
- id: S10
  title: NVIDIA MIG User Guide
  url: https://docs.nvidia.com/datacenter/tesla/mig-user-guide/
  summary: GPU instances, compute instances, profiles, and partitioning behavior.
- id: S11
  title: NVIDIA NCCL collectives documentation
  url: https://docs.nvidia.com/deeplearning/nccl/user-guide/docs/usage/collectives.html
  summary: AllReduce, ReduceScatter, AllGather, Broadcast, Reduce, point-to-point, and collective
    communication semantics.
- id: S12
  title: PyTorch DistributedDataParallel docs
  url: https://pytorch.org/docs/stable/generated/torch.nn.parallel.DistributedDataParallel.html
  summary: distributed training synchronizes gradients across model replicas.
- id: S13
  title: PyTorch FullyShardedDataParallel docs
  url: https://pytorch.org/docs/stable/fsdp.html
  summary: parameter sharding and collective communication such as all-gather and reduce-scatter during
    training.
- id: S14
  title: DeepSpeed ZeRO-3 docs
  url: https://deepspeed.readthedocs.io/en/stable/zero3.html
  summary: ZeRO partitions optimizer states, gradients, and parameters, requiring coordinated all-gather/reduce-scatter
    style movement.
- id: S15
  title: JAX Scaling Book, training parallelism
  url: https://jax-ml.github.io/scaling-book/training/
  summary: mechanism-level discussion of data, tensor, pipeline, and sharded training communication
    costs.
- id: S16
  title: Demystifying the Communication Characteristics for Distributed Transformer Models
  url: https://arxiv.org/abs/2408.10197
  summary: research evidence that transformer training produces structured communication traces and
    that communication can dominate at scale.
- id: S17
  title: A Tale of Two Cs, Computation vs. Communication Scaling
  url: https://research.cs.wisc.edu/hal/papers/spati-iiswc23-totc.pdf
  summary: research on data-parallel and tensor-parallel communication/computation scaling.
- id: S18
  title: Slurm `sacct` docs
  url: https://slurm.schedmd.com/sacct.html
  summary: accounting records for jobs, elapsed time, states, allocations, TRES, energy, and selected
    I/O.
- id: S19
  title: Slurm GRES docs
  url: https://slurm.schedmd.com/gres.html
  summary: GPU generic resources, typed GPUs, NVML autodetection, MIG/MPS configuration, and GPU accounting
    surfaces.
- id: S20
  title: Kubernetes device plugins
  url: https://kubernetes.io/docs/concepts/extend-kubernetes/compute-storage-net/device-plugins/
  summary: vendor devices exposed as extended resources and allocated to pods.
- id: S21
  title: Kubernetes audit logging
  url: https://kubernetes.io/docs/tasks/debug/debug-cluster/audit/
  summary: chronological API-server audit records under configured policy.
- id: S22
  title: Kubernetes Events API
  url: https://kubernetes.io/docs/reference/kubernetes-api/cluster-resources/event-v1/
  summary: source-emitted lifecycle events with involved objects, reasons, and event times.
- id: S23
  title: AWS EC2 Capacity Blocks for ML
  url: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ec2-capacity-blocks.html
  summary: capacity reservation start time, duration, instance type, and quantity concepts for accelerated
    instances.
- id: S24
  title: AWS EC2 P5/P5e/P5en instance pages
  url: https://aws.amazon.com/ec2/instance-types/p5/
  summary: H100/H200 instances, accelerator count, EFA, and high-throughput networking context.
- id: S25
  title: AWS Elastic Fabric Adapter
  url: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa.html
  summary: low-latency, high-throughput network interface for tightly coupled HPC and ML workloads.
- id: S26
  title: AWS Cost and Usage Reports
  url: https://docs.aws.amazon.com/cur/latest/userguide/what-is-cur.html
  summary: usage line items, usage type, operation, time interval, product, and resource fields when
    enabled.
- id: S27
  title: AWS CloudTrail concepts
  url: https://docs.aws.amazon.com/awscloudtrail/latest/userguide/cloudtrail-concepts.html
  summary: management, data, network activity, event time, resource, and actor records, with coverage
    depending on configuration.
- id: S28
  title: AWS VPC Flow Logs records
  url: https://docs.aws.amazon.com/vpc/latest/userguide/flow-log-records.html
  summary: byte/packet records, aggregation intervals, traffic-path fields, and skipped-data/delivery
    caveats.
- id: S29
  title: Google Cloud VPC Flow Logs
  url: https://cloud.google.com/vpc/docs/flow-logs
  summary: sampled and aggregated flow logs, metadata annotations, aggregation intervals, and RDMA
    flow caveats.
- id: S30
  title: Azure NSG Flow Logs overview
  url: https://learn.microsoft.com/en-us/azure/network-watcher/nsg-flow-logs-overview
  summary: inbound/outbound flow logging and collection/export limitations.
- id: S31
  title: Amazon S3 server access logging
  url: https://docs.aws.amazon.com/AmazonS3/latest/userguide/ServerLogs.html
  summary: object request logs, operation, bytes, timing, and best-effort delivery caveats.
- id: S32
  title: Amazon S3 log format
  url: https://docs.aws.amazon.com/AmazonS3/latest/userguide/LogFormat.html
  summary: fields for operation, object key, bytes sent, object size, requester, and timing.
- id: S33
  title: Amazon FSx for Lustre metrics
  url: https://docs.aws.amazon.com/fsx/latest/LustreGuide/fs-metrics.html
  summary: filesystem throughput, metadata operations, client I/O, queueing, and storage performance
    metrics.
- id: S34
  title: Google Cloud Audit Logs
  url: https://cloud.google.com/logging/docs/audit
  summary: Admin Activity, Data Access, System Event, and Policy Denied audit log types.
- id: S35
  title: Google Cloud detailed billing export
  url: https://cloud.google.com/billing/docs/how-to/export-data-bigquery-tables/detailed-usage
  summary: usage, SKU, project, service, labels, and resource fields in BigQuery export.
- id: S36
  title: Google Compute Engine GPU monitoring
  url: https://cloud.google.com/compute/docs/gpus/monitor-gpus
  summary: GPU utilization and metric collection for Compute Engine VMs.
- id: S37
  title: Google Cloud TPU monitoring library
  url: https://cloud.google.com/tpu/docs/tpu-monitoring-library
  summary: TPU TensorCore utilization, duty cycle, memory, and related metrics.
- id: S38
  title: AWS Container Insights enhanced observability metrics
  url: https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/Container-Insights-metrics-enhanced-EKS.html
  summary: EKS/container, GPU, Neuron, EFA, pod, node, and network/storage infrastructure metrics.
- id: S39
  title: AWS Neuron monitoring docs
  url: https://awsdocs-neuron.readthedocs-hosted.com/en/latest/tools/neuron-sys-tools/neuron-monitor-user-guide.html
  summary: accelerator utilization, memory, runtime, and Neuron-device metrics.
- id: S40
  title: DMTF Redfish standards
  url: https://www.dmtf.org/standards/redfish
  summary: standard management model for servers, power, thermal, telemetry, logs, fabrics, storage,
    firmware, and cooling equipment.
- id: S41
  title: Redfish Resource and Schema Guide DSP2046
  url: https://redfish.dmtf.org/schemas/DSP2046_2025.2.html
  summary: schemas for chassis, power, thermal, environment metrics, power distribution, log services,
    firmware, and related management records.
- id: S42
  title: NVIDIA UFM Telemetry docs
  url: https://docs.nvidia.com/networking/display/ufmsdnappumv4160/telemetry
  summary: InfiniBand fabric telemetry, bandwidth, latency, congestion, errors, and XmitWait-style
    fabric counters.
- id: S43
  title: NVIDIA UFM supported port counters/events
  url: https://docs.nvidia.com/networking/display/ufmsdnappumv41814/Appendix+-+Supported+Port+Counters+and+Events
  summary: transmit/receive counters, congestion, errors, and port event surfaces.
- id: S44
  title: Epoch AI data centers dataset
  url: https://epoch.ai/data/data-centers
  summary: public site-specific power, accelerator, commissioning, and public-evidence context for
    AI datacenters.
- id: S45
  title: PyTorch Distributed Checkpoint docs
  url: https://docs.pytorch.org/docs/main/distributed.checkpoint.html
  summary: distributed checkpoint save/load, per-rank checkpoint files, and topology-independent resharding
    support.
- id: S46
  title: DeepSpeed model checkpointing docs
  url: https://deepspeed.readthedocs.io/en/latest/model-checkpointing.html
  summary: training checkpoint save/load, optimizer/scheduler state handling, and ZeRO checkpoint
    behavior.
- id: S47
  title: TensorFlow training checkpoint guide
  url: https://www.tensorflow.org/guide/checkpoint
  summary: periodic checkpoint writes, optimizer/model state, and checkpoint file layout.